Andreas Duus Pape
Assistant Professor of Economics
Binghamton University (SUNY)
Kenneth Kurtz
Associate Professor of Psychology
Binghamton University (SUNY)
Is Case-based Decision Theory consistent with Empirical Patterns of Human Classification Learning?
Monday, February 7, 2011
Science 1 room 149, 5:00 PM
Abstract
We present a computational implementation of Case-based Decision Theory (CBDT) (Gilboa & Schmeidler, 1995) that can be used for agents in an arbitrary choice problem or game, so can be used to generate ’empirical’ choice behavior of a CBDT-governed agent with regard to a finite problem or game. We use this implementation to test the performance of CBDT on a benchmark problem from the psychological literature on human classification learning (Nosofsky, Gluck, Palmeri, & McKinley, 1994; Shepard, Hovland, & Jenkins, 1961). We evaluate whether the behavior of a CBDT-governed agent is consistent with empirically established patterns of human cognition.
Biographies
Dr. Pape received a PhD in economics from the University of Michigan in 2007 under Emre Ozdenoren and Scott E. Page. He has two primary veins of research. The first is an investigation of agents as model-makers in decision theory. This includes work representing agents’ causal models of problems in a decision-theoretic setting and developing software which represents the decision-making of an agent governed by case-based decision theory in order to empirically test the validity of that decision theory (work with Dr. Kurtz). The other vein of research is game-theoretic and agent-based modeling in support of applied environmental and public finance questions; one notable example is the rational-expectations model of agents in a community voting over local revenue levels under uncertain tax prices used in the paper with Anderson about limits on local property taxes, another is an agent-based model of the physics of groundwater for optimal water resource planning problems.
Dr. Kurtz earned his Ph.D at Stanford University under the direction of Dr. David Rumelhart in 1997. He then completed post-doctoral work in the study of learning and reasoning by analogy with Dr. Dedre Gentner at Northwestern University. Dr. Kurtz joined the Department of Psychology at Binghamton University in 2001 where he now directs the Learning and Representation in Cognition (LaRC) Laboratory. Dr. Kurtz pursues research on a range of topics in cognitive science including computational models of category learning, concepts and semantic knowledge, machine learning, cognition and student learning, analogy, similarity, and creativity.
Assigned Reading
- Nosofsky, R. M., and Palmeri, T. J. “Learning to classify integral-dimension stimuli.” Pscyhonomic Bulletin & Review 1996, 3 (2). [PDF; “This is a psychology paper which describes the background of the experiment we run with this algorithm, to test the algorithm’s ‘humanness.'”]
Additional Readings
- Gilboa, I., and Schmeidler, D. “Case-Based Decision Theory.” The Quarterly Journal of Economics 1995, Vol. 110, No. 3. [PDF; “This is an economics Decision Theory paper that describes the algorithm we implement.”]
